Automatic Speaker-Identification

Starting 2024, we’re introducing a new feature for podcasters: automatic speaker identification in podcast episodes.

Until now, identifying speakers was a manual task in the PodcastAI workflow, requiring users to scroll through transcripts to identify each speaker. This process, particularly in episodes with several speakers, could take up to a couple of minutes.

Now, you just hit auto-identify, which quickly gives you auto-identifications in seconds. Hosts are automatically identified, as you might expect. But the real gem here is the auto-identification of guests via AI — it can deduce that from the conversation alone.

The host auto-identification can be accepted or rejected, but the guest can also be corrected.

You can then accept or reject each auto-identification in a click. There is one potential issue: if the guest name was incorrectly transcribed (which can happen with names, since they have a tendency to be irregular), that will be the suggested guest name. But we thought of that — so you’re offered a third option, which lets you correct the guest name.

The best part of this is that we now have what we needed to implement a full-auto feature in the future. It won’t be for everyone, but podcasters who want full-auto will be able to simply drop in their new episode and walk-away in the knowledge that every post-production task will automatically engage in the appropriate sequence, resulting in their fully post-produced episode reaching every podcast platform.

This is just one more step in the core mission of PodcastAI: automating your podcast. We have lots more in store — 2024 is going to be incredible.